Stop Measuring Things That Do Not Matter

Most dashboards are just vanity mirrors for engineering teams. If a number doesn’t help you decide whether to ship a feature or roll back a deploy, it’s just noise.

What makes a metric useful?

A good metric is actionable. If I see that my database latency increased by 10ms, I need to know exactly which index or query is to blame. If I’m looking at “total daily active users” on a dashboard that doesn’t tie back to a recent deployment, I’m just watching a line move. That’s not engineering. That’s just data theater.

Here is how I track real performance in a Node.js context:

1
// Instead of tracking total reqs, track duration by endpoint
2
const start = Date.now();
3
await processRequest();
4
const duration = Date.now() - start;
5

6
metrics.histogram('api_request_duration_ms', duration, { endpoint: '/login' });

By tagging the endpoint, I can see if a specific refactor caused a regression. Without the tag, the metric is useless (I suppose it’s fine for high-level marketing reports, but I don’t write code for those).

“But Namir, surely high-level business metrics matter for team alignment?” I hear you. You’re right that leadership needs visibility, but engineering teams often confuse business KPIs with technical health. They aren’t the same. If your business metric moves, you rarely have enough context to fix it without looking at the underlying technical signals anyway. Focus on the signals that actually change your behavior.

Sometimes I track things just to satisfy curiosity, but I try to delete them after a week. If I’m not looking at the graph to solve a problem, it’s just cluttering my monitor (and my head).

What’s the one metric you check every day that actually dictates your next code change?